An intelligent control method for power distribution station based on deep integration of digital and analog
Through digital-to-analog deep fusion technology, a three-dimensional digital model and anomaly detection model of distribution station buildings are built, and combined with SLAM algorithm, intelligent management of distribution station buildings is realized, solving the information asymmetry and inefficiency in the traditional management model. Problem, improving management efficiency and security.
Patent Information
- Application Number
- CN202411179239.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-27
AI Technical Summary
The traditional distribution station management model has information asymmetry, inefficiency and safety risks, and cannot meet the needs of modern power systems.
Using intelligent management and control methods based on digital-analog deep fusion, by arranging multiple sensors in the distribution station building, collecting data in real time and building a three-dimensional digital model, combining deep learning and SLAM algorithms, the abnormal detection and autonomous navigation of patrol robots are realized, and the inspection robots are optimized. Inspection path and equipment management.
It realizes intelligent and automated management of distribution station buildings, improves inspection efficiency and timely processing of equipment status, and enhances management reliability and safety.
Smart Images

Figure CN118691266B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power distribution station management, and in particular to an intelligent control method for power distribution station based on deep integration of digital and analog. Background Art
[0002] With the rapid development of the power industry and the in-depth advancement of intelligent transformation, the distribution station, as a core component of the distribution network system, its operating efficiency, safety and management level directly affect the reliability and stability of the power supply of the entire distribution network.
[0003] With the development of intelligent distribution network and Internet of Things technology, higher and higher requirements are put forward for the operation and management of distribution substations. The traditional distribution substation management mode often has problems such as information asymmetry, low efficiency, and safety hazards, which cannot meet the needs of modern power systems. Summary of the invention
[0004] In order to solve the above problems, the purpose of the present invention is to provide a method for intelligent management and control of distribution stations based on deep integration of digital and analog, which realizes the full process automation and intelligence of intelligent management and control of distribution stations through deep integration of digital and analog.
[0005] To achieve the above object, the present invention adopts the following technical solutions:
[0006] A method for intelligent control of power distribution stations based on deep integration of digital and analog, comprising the following steps:
[0007] S1: Multiple types of sensors are arranged in the distribution station, including temperature and humidity sensors, voltage and current sensors, vibration sensors, and gas sensors. IoT gateways and edge computing devices are used to collect sensor data in real time and transmit it to the data processing center;
[0008] S2: The data processing center preprocesses the sensor data, including removing noise data, filling missing values, standardizing data formats, and time aligning and synchronizing data with different sampling frequencies and timestamps;
[0009] S3: construct a three-dimensional digital model of the power distribution station, and in the digital model, accurately identify the location and type of each device, assign a unique identifier, and associate the collected sensor data with the corresponding digital model device through the ID;
[0010] S4: Build an anomaly detection model based on the deep learning model, and detect abnormal devices based on the sensor data collected in real time;
[0011] S5: Based on the three-dimensional digital model of the power distribution station, combined with the station layout and equipment importance, the shortest path algorithm is used to plan the optimal path of the inspection robot, and the inspection path is adaptively adjusted according to the equipment abnormality detection results;
[0012] S6: Based on path planning, the SLAM algorithm is used in combination with the three-dimensional digital model of the distribution station to realize the autonomous positioning and navigation of the inspection robot.
[0013] Furthermore, step S3 is specifically as follows:
[0014] Use a 3D laser scanner to fully scan the power distribution station to obtain 3D point cloud data inside and outside the power distribution station, including the location, shape and size of the equipment;
[0015] Combine high-definition cameras and panoramic cameras to obtain high-definition images of the substation room, external equipment and environment for texture mapping of the model;
[0016] The collected point cloud data is subjected to noise reduction and filtering to remove errors and noise points, and the point cloud data of each scanning position is aligned and merged using the ICP algorithm to construct a complete 3D point cloud model;
[0017] ;
[0018] in, and are the i-th point of the source point cloud and the target point cloud respectively; R is the rotation matrix, and t is the translation vector; Represents the Euclidean norm of the solution vector;
[0019] Use the Poisson Surface Reconstruction algorithm to convert the processed point cloud data into a triangular mesh; and map the high-definition image texture onto the triangular mesh surface to improve the visual effect of the model;
[0020] Based on the device recognition algorithm, the scanned devices are classified and identified, using image and point cloud data as input, and the model outputs the device type;
[0021] Based on the 3D point cloud, calculate the physical space position of each device:
[0022] ;
[0023] Among them, P local It is three-dimensional point cloud data;
[0024] Assign a uuid to each device to form a unique identifier to facilitate subsequent data management and query;
[0025] Use 3D modeling software and real-time rendering engines to import digital models for real-time rendering and display;
[0026] The three-dimensional digital model, location information and unique identifier of the equipment are stored in the database to manage the equipment information. With the help of the Internet of Things platform, the data collected by the sensor is bound to the digital model to achieve real-time data display and equipment status update.
[0027] Furthermore, the processed point cloud data is converted into a triangular mesh using the Poisson Surface Reconstruction algorithm, as follows:
[0028] The normal vector is calculated for each processed point cloud data, and a smooth surface is generated by solving the Poisson equation based on the Poisson equation. The Poisson equation is as follows:
[0029] ;
[0030] in, is the vector differential operator, V is the gradient of the implicit function to be solved, and N is the normal vector of each point in the point cloud;
[0031] Use octree to segment the point cloud, build a weight field on the octree node to represent the influence distribution of each point cloud point, discretize the Poisson equation, and build a linear system based on the octree node:
[0032] ;
[0033] Among them, A is the coefficient matrix, which represents the relationship between octree nodes; x is the potential value of the octree node; b is the weight vector, which represents the normal vector field of the point cloud point;
[0034] The conjugate gradient method is used to solve the linear equations to obtain the potential value of each node;
[0035] According to the potential value obtained by solving, isosurface extraction is performed to generate a triangular mesh surface.
[0036] Furthermore, the conjugate gradient method is used to solve the linear equations to obtain the potential value of each node, as follows:
[0037] Suppose the initial potential value vector x0 is set to the zero vector: x0=0, calculate the initial residual r0 and the initial search direction p0:
[0038]
[0039] Use the conjugate gradient method to iterate and update until convergence. For the kth iteration:
[0040] ①Calculation step length:
[0041]
[0042] Among them, the superscript T represents transposition;
[0043] ② Update the potential value vector and residual :
[0044] ;
[0045] ;
[0046] ③Calculate the scale factor :
[0047] ;
[0048] ④Update search direction :
[0049] ;
[0050] The iteration stops when the residual norm is less than a given threshold or the maximum number of iterations is reached.
[0051] Furthermore, with the help of the IoT platform, the data collected by the sensor is bound to the digital model to achieve real-time data display and device status update, as follows: use MQTT, HTTP or CoAP protocol to upload sensor data to the IoT platform; the IoT platform processes and stores the data uploaded by the sensor and binds it according to the device ID; the terminal manages the device through the IoT platform; and through the IoT platform, collects real-time data from the sensor and transmits it to the front-end application for display.
[0052] Furthermore, step S4 is specifically as follows:
[0053] Building anomaly detection based on a deep autoencoder, wherein the deep autoencoder includes an encoder and a decoder;
[0054] Each layer of the encoder is processed by the activation function ReLU, and compressed layer by layer to a low-dimensional representation:
[0055] ;
[0056] Among them, Z is the encoding result, is the encoder function; X is the input data; is a set of encoder parameters;
[0057] The decoder reconstructs the original dimension layer by layer from the low-dimensional representation:
[0058] ;
[0059] in, is the decoding result; Decoder function; is a set of decoder parameters;
[0060] A regularization term is added to the loss function to prevent overfitting, and the improved loss function is as follows:
[0061] ;
[0062] in, is the total loss; is the reconstruction error; and They are the L1 regularization coefficient and the L2 regularization coefficient respectively; is the L1 norm of the model parameters; is the L2 norm of the model parameters;
[0063] During anomaly detection, the reconstruction error of each data point is calculated. If the error exceeds the dynamically adjusted threshold, it is determined to be an anomaly.
[0064] Furthermore, S5 is specifically:
[0065] S51: Based on the three-dimensional digital model of the distribution station, construct a distribution station directed graph G=( ): It represents all locations in the station, including equipment locations and intermediate locations that are passed by during inspections; E represents the path between nodes;
[0066] Each edge There is a weight , indicating a node Calculate the distance between
[0067] S52: Use Dijkstra algorithm to calculate the shortest path;
[0068] S53: Dynamically adjust the inspection path using equipment abnormalities and importance levels, by introducing an adaptive adjustment factor into the shortest path algorithm:
[0069] When the device is normal, the path weight is the distance plus the importance adjustment factor:
[0070] ;
[0071] If the node If an exception occurs, the path weight will increase by an exception weight. :
[0072] ;
[0073] in, Representation Node The physical distance between Representation Node level of importance; Representation Node Abnormal situation, 0 means no abnormality, 1 means abnormality; is the importance modifier; It is an abnormal regulatory factor.
[0074] Furthermore, the Dijkstra algorithm is used to calculate the shortest path, as follows:
[0075] Set the initial distance for all nodes v , set dist(s)=0 for the starting node s, create a priority queue, and add the starting node to the queue;
[0076] Extract the node u with the smallest distance from the priority queue. If node u has not been processed yet, mark it as processed;
[0077] For Node , For all neighbor nodes of node u, calculate the distance from the starting node s to the node The path distance dist( ), if the path through u is shorter, update dist( ) and add it to the priority queue;
[0078] Repeat the above steps until all nodes have been processed or the target node has been reached.
[0079] Furthermore, step S6 is specifically as follows:
[0080] S61: Real-time acquisition of environmental data through the laser radar, IMU and camera configured on the inspection robot;
[0081] S62: Utilize SLAM algorithm to continuously update the three-dimensional map of the power distribution station during navigation;
[0082] S63: During the path planning process, the known position of the 3D model and the map constructed by SLAM in real time are combined to correct the current position of the robot and adjust the inspection path in real time.
[0083] Furthermore, the SLAM algorithm is used to continuously update the three-dimensional map of the power distribution station during the navigation process, as follows:
[0084] The robot state is integrated into several particles, each particle includes its own position state and the corresponding map;
[0085] According to the control input u t Make state predictions for each particle and add noise:
[0086]
[0087] in, Represents particles In time Status; Indicates at time The control input of f() is the motion model function; Represents particles In time The weight of
[0088] Update the map information in each particle and calculate the likelihood between the map and the observation:
[0089] Resample according to the likelihood value and keep the most likely particles:
[0090] ;
[0091] ;
[0092] in, Represents particles In time Maps; Indicates that in the particle state and map Next, observe The probability of Represents resampling.
[0093] The present invention has the following beneficial effects:
[0094] 1. The present invention realizes the full process automation and intelligence of intelligent management and control of power distribution stations through deep integration of digital and analog;
[0095] 2. The present invention combines point cloud data processing and 3D modeling technology to construct a high-precision 3D model of the power distribution station, and then uses high-definition image data for texture mapping, ultimately achieving refined expression of the model and enhanced visual effects. The conjugate gradient method is used to effectively solve the linear equations in Poisson surface reconstruction, and can process large-scale sparse matrices, achieve rapid convergence, and obtain high-quality 3D surface models.
[0096] 3. The present invention combines the layout of the station house and the importance of the equipment, uses the shortest path algorithm to plan the optimal path of the inspection robot, and according to the abnormal detection results of the equipment, adaptively adjusts the inspection path, prioritizes the inspection of abnormal equipment, improves the inspection efficiency, and adaptively adjusts to ensure that abnormal equipment is inspected and processed in time. And further based on the path planning, combined with the three-dimensional digital model of the power distribution station, the SLAM algorithm is used to realize the autonomous positioning and navigation of the inspection robot. BRIEF DESCRIPTION OF THE DRAWINGS
[0097] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0098] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:
[0099] refer to Figure 1 In this implementation, a method for intelligent control of power distribution stations based on deep integration of digital and analog is provided, comprising the following steps:
[0100] S1: Multiple types of sensors are arranged in the distribution station, including temperature and humidity sensors, voltage and current sensors, vibration sensors, and gas sensors. IoT gateways and edge computing devices are used to collect sensor data in real time and transmit it to the data processing center;
[0101] S2: The data processing center preprocesses the sensor data, including removing noise data, filling missing values, standardizing data formats, and time aligning and synchronizing data with different sampling frequencies and timestamps;
[0102] S3: construct a three-dimensional digital model of the power distribution station, and in the digital model, accurately identify the location and type of each device, assign a unique identifier, and associate the collected sensor data with the corresponding digital model device through the ID;
[0103] S4: Build an anomaly detection model based on the deep learning model, and detect abnormal devices based on the sensor data collected in real time;
[0104] S5: Based on the three-dimensional digital model of the power distribution station, combined with the station layout and equipment importance, the shortest path algorithm is used to plan the optimal path of the inspection robot, and the inspection path is adaptively adjusted according to the equipment abnormality detection results;
[0105] S6: Based on path planning, the SLAM algorithm is used in combination with the three-dimensional digital model of the distribution station to realize the autonomous positioning and navigation of the inspection robot.
[0106] In this implementation, step S3 is specifically as follows:
[0107] Use a 3D laser scanner to fully scan the power distribution station to obtain 3D point cloud data inside and outside the power distribution station, including the location, shape and size of the equipment;
[0108] Combine high-definition cameras and panoramic cameras to obtain high-definition images of the substation room, external equipment and environment for texture mapping of the model;
[0109] The collected point cloud data is subjected to noise reduction and filtering to remove errors and noise points, and the point cloud data of each scanning position is aligned and merged using the ICP algorithm to construct a complete 3D point cloud model;
[0110] ;
[0111] in, and are the i-th point of the source point cloud and the target point cloud respectively; R is the rotation matrix, and t is the translation vector; Represents the Euclidean norm of the solution vector;
[0112] Use the Poisson Surface Reconstruction algorithm to convert the processed point cloud data into a triangular mesh; and map the high-definition image texture onto the triangular mesh surface to improve the visual effect of the model;
[0113] Based on device recognition algorithms (such as convolutional neural networks (CNNs)), the scanned devices are classified and identified, using images and point cloud data as input, and the model outputs the device type;
[0114] Based on the 3D point cloud, calculate the physical space position of each device:
[0115] ;
[0116] Among them, P local It is three-dimensional point cloud data;
[0117] Assign a uuid to each device to form a unique identifier to facilitate subsequent data management and query;
[0118] Use 3D modeling software and real-time rendering engines to import digital models for real-time rendering and display;
[0119] The three-dimensional digital model, location information and unique identifier of the equipment are stored in the database to manage the equipment information. With the help of the Internet of Things platform, the data collected by the sensor is bound to the digital model to achieve real-time data display and equipment status update.
[0120] In this implementation, the Poisson Surface Reconstruction algorithm is used to convert the processed point cloud data into a triangular mesh as follows:
[0121] The normal vector is calculated for each processed point cloud data, and a smooth surface is generated by solving the Poisson equation based on the Poisson equation. The Poisson equation is as follows:
[0122] ;
[0123] in, is the vector differential operator, V is the gradient of the implicit function to be solved, and N is the normal vector of each point in the point cloud;
[0124] Use octree to segment the point cloud, build a weight field on the octree node to represent the influence distribution of each point cloud point, discretize the Poisson equation, and build a linear system based on the octree node:
[0125] ;
[0126] Among them, A is the coefficient matrix, which represents the relationship between octree nodes; x is the potential value of the octree node; b is the weight vector, which represents the normal vector field of the point cloud point;
[0127] The conjugate gradient method is used to solve the linear equations to obtain the potential value of each node;
[0128] According to the potential value obtained by solving, isosurface extraction is performed to generate a triangular mesh surface.
[0129] In this implementation, the conjugate gradient method is used to solve the linear equations to obtain the potential value of each node, as follows:
[0130] Suppose the initial potential value vector x0 is set to the zero vector: x0=0, calculate the initial residual r0 and the initial search direction p0:
[0131]
[0132] Use the conjugate gradient method to iterate and update until convergence. For the kth iteration:
[0133] ①Calculation step length:
[0134]
[0135] Among them, the superscript T represents transposition;
[0136] ② Update the potential value vector and residual :
[0137] ;
[0138] ;
[0139] ③Calculate the scale factor :
[0140] ;
[0141] ④Update search direction :
[0142] ;
[0143] The iteration stops when the residual norm is less than a given threshold or the maximum number of iterations is reached.
[0144] In this implementation, with the help of the Internet of Things platform, the data collected by the sensor is bound to the digital model to achieve real-time data display and device status update, as follows: the sensor data is uploaded to the Internet of Things platform using the MQTT, HTTP or CoAP protocol; the Internet of Things platform processes and stores the data uploaded by the sensor and binds it according to the device ID; the terminal (such as mobile devices, desktop applications) manages the device through the Internet of Things platform (such as remote update); and the real-time data of the sensor is collected through the Internet of Things platform and transmitted to the front-end application for display.
[0145] In this implementation, step S4 is specifically as follows:
[0146] Building anomaly detection based on a deep autoencoder, wherein the deep autoencoder includes an encoder and a decoder;
[0147] Each layer of the encoder is processed by the activation function ReLU, and compressed layer by layer to a low-dimensional representation:
[0148] ;
[0149] Among them, Z is the encoding result, is the encoder function; X is the input data; is a set of encoder parameters;
[0150] The decoder reconstructs the original dimension layer by layer from the low-dimensional representation:
[0151] ;
[0152] in, is the decoding result; Decoder function; is a set of decoder parameters;
[0153] A regularization term is added to the loss function to prevent overfitting, and the improved loss function is as follows:
[0154] ;
[0155] in, is the total loss; is the reconstruction error; and They are the L1 regularization coefficient and the L2 regularization coefficient respectively; is the L1 norm of the model parameters; is the L2 norm of the model parameters;
[0156] During anomaly detection, the reconstruction error of each data point is calculated. If the error exceeds the dynamically adjusted threshold, it is determined to be an anomaly.
[0157] In this implementation, S5 is specifically:
[0158] S51: Based on the three-dimensional digital model of the distribution station, construct a distribution station directed graph G=( ): It represents all locations in the station, including equipment locations and intermediate locations that are passed by during inspections; E represents the path between nodes;
[0159] Each edge There is a weight , indicating a node Calculate the distance between
[0160] S52: Use Dijkstra algorithm to calculate the shortest path;
[0161] S53: Dynamically adjust the inspection path using equipment abnormalities and importance levels, by introducing an adaptive adjustment factor into the shortest path algorithm:
[0162] When the device is normal, the path weight is the distance plus the importance adjustment factor:
[0163] ;
[0164] If the node If an exception occurs, the path weight will increase by an exception weight. :
[0165] ;
[0166] in, Representation Node The physical distance between Representation Node level of importance; Representation Node Abnormal situation, 0 means no abnormality, 1 means abnormality; is the importance modifier; It is an abnormal regulatory factor.
[0167] In this implementation, the Dijkstra algorithm is used to calculate the shortest path, as follows:
[0168] Set the initial distance for all nodes v , set dist(s)=0 for the starting node s, create a priority queue, and add the starting node to the queue;
[0169] Extract the node u with the smallest distance from the priority queue. If node u has not been processed yet, mark it as processed;
[0170] For Node , For all neighbor nodes of node u, calculate the distance from the starting node s to the node The path distance dist( ), if the path through u is shorter, update dist( ) and add it to the priority queue;
[0171] Repeat the above steps until all nodes have been processed or the target node has been reached.
[0172] In this implementation, step S6 is specifically as follows:
[0173] S61: Real-time acquisition of environmental data through the laser radar, IMU and camera configured on the inspection robot;
[0174] S62: Utilize SLAM algorithm to continuously update the three-dimensional map of the power distribution station during navigation;
[0175] S63: During the path planning process, the known position of the 3D model and the map constructed by SLAM in real time are combined to correct the current position of the robot and adjust the inspection path in real time.
[0176] In this implementation, the SLAM algorithm is used to continuously update the three-dimensional map of the power distribution station during the navigation process, as follows:
[0177] The robot state is integrated into several particles, each particle includes its own position state and the corresponding map;
[0178] According to the control input u t Make state predictions for each particle and add noise:
[0179]
[0180] in, Represents particles In time Status; Indicates at time The control input of f() is the motion model function; Represents particles In time The weight of
[0181] Update the map information in each particle and calculate the likelihood between the map and the observation:
[0182] Resample according to the likelihood value and keep the most likely particles:
[0183] ;
[0184] ;
[0185] in, Represents particles In time Maps; Indicates that in the particle state and map Next, observe The probability of Represents resampling.
[0186] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0187] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0188] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0189] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0190] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.
Claims
1. An intelligent control method for power distribution stations based on deep integration of digital and analog, It is characterized in that The following steps are involved: S1: Multiple types of sensors are arranged in the distribution station, including temperature and humidity sensors, voltage and current sensors, vibration sensors, and gas sensors. IoT gateways and edge computing devices are used to collect sensor data in real time and transmit it to the data processing center; S2: The data processing center preprocesses the sensor data, including removing noise data, filling missing values, standardizing data formats, and time aligning and synchronizing data with different sampling frequencies and timestamps; S3: construct a three-dimensional digital model of the power distribution station, and in the digital model, accurately identify the location and type of each device, assign a unique identifier, and associate the collected sensor data with the corresponding digital model device through the ID; S4: Build an anomaly detection model based on the deep learning model, and detect abnormal devices based on the sensor data collected in real time; S5: Based on the three-dimensional digital model of the power distribution station, combined with the station layout and equipment importance, the shortest path algorithm is used to plan the optimal path of the inspection robot, and the inspection path is adaptively adjusted according to the equipment abnormality detection results; S6: Based on path planning, the SLAM algorithm is used in combination with the three-dimensional digital model of the power distribution station to realize the autonomous positioning and navigation of the inspection robot; The step S3 is specifically as follows: Use a 3D laser scanner to fully scan the power distribution station to obtain 3D point cloud data inside and outside the power distribution station, including the location, shape and size of the equipment; Combine high-definition cameras and panoramic cameras to obtain high-definition images of the substation room, external equipment and environment for texture mapping of the model; The collected point cloud data is subjected to noise reduction and filtering to remove errors and noise points, and the point cloud data of each scanning position is aligned and merged using the ICP algorithm to construct a complete 3D point cloud model; ; in, and are the i-th point of the source point cloud and the target point cloud respectively; R is the rotation matrix, and t is the translation vector; Represents the Euclidean norm of the solution vector; Use the Poisson Surface Reconstruction algorithm to convert the processed point cloud data into a triangular mesh; and map the high-definition image texture onto the triangular mesh surface to improve the visual effect of the model; Based on the device recognition algorithm, the scanned devices are classified and identified, using image and point cloud data as input, and the model outputs the device type; Based on the 3D point cloud, calculate the physical space position of each device: ; Among them, P local It is three-dimensional point cloud data; Assign a uuid to each device to form a unique identifier to facilitate subsequent data management and query; Use 3D modeling software and real-time rendering engines to import digital models for real-time rendering and display; The three-dimensional digital model, location information and unique identifier of the equipment are stored in the database to manage the equipment information. With the help of the Internet of Things platform, the data collected by the sensor is bound to the digital model to achieve real-time data display and equipment status update.
2. According to claim 1, a method for intelligent management and control of power distribution stations based on deep fusion of digital and analog, It is characterized in that The Poisson Surface Reconstruction algorithm is used to convert the processed point cloud data into a triangular mesh, as follows: The normal vector is calculated for each processed point cloud data, and a smooth surface is generated by solving the Poisson equation based on the Poisson equation. The Poisson equation is as follows: ; in, is the vector differential operator, V is the gradient of the implicit function to be solved, and N is the normal vector of each point in the point cloud; Use octree to segment the point cloud, build a weight field on the octree node to represent the influence distribution of each point cloud point, discretize the Poisson equation, and build a linear system based on the octree node: ; Among them, A is the coefficient matrix, which represents the relationship between octree nodes; x is the potential value of the octree node; b is the weight vector, which represents the normal vector field of the point cloud point; The conjugate gradient method is used to solve the linear equations to obtain the potential value of each node; According to the potential value obtained by solving, isosurface extraction is performed to generate a triangular mesh surface.
3. According to the method for intelligent control of power distribution station based on deep fusion of digital and analog according to claim 2, It is characterized in that The conjugate gradient method is used to solve the linear equations to obtain the potential value of each node, as follows: Let the initial potential value vector x 0 Set to zero vector: x 0 =0, calculate the initial residual r 0 and the initial search direction p 0 : Use the conjugate gradient method to iterate and update until convergence. For the kth iteration: ①Calculation step length: Among them, the superscript T represents transposition; ② Update the potential value vector and residual : ; ; ③Calculate the scale factor : ; ④Update search direction : ; The iteration stops when the residual norm is less than a given threshold or the maximum number of iterations is reached.
4. According to claim 1, a method for intelligent management and control of power distribution stations based on deep fusion of digital and analog, It is characterized in that The IoT platform is used to bind the data collected by the sensor to the digital model to achieve real-time data display and device status update, specifically as follows: the sensor data is uploaded to the IoT platform using MQTT, HTTP or CoAP protocol; the IoT platform processes and stores the data uploaded by the sensor and binds it according to the device ID; The terminal manages the equipment through the IoT platform; And through the Internet of Things platform, real-time data from sensors is collected and transmitted to the front-end application for display.
5. According to claim 1, a method for intelligent control of power distribution stations based on deep fusion of digital and analog, It is characterized in that The step S4 is specifically as follows: Building anomaly detection based on a deep autoencoder, wherein the deep autoencoder includes an encoder and a decoder; Each layer of the encoder is processed by the activation function ReLU, and compressed layer by layer to a low-dimensional representation: ; Among them, Z is the encoding result, is the encoder function; X is the input data; is a set of encoder parameters; The decoder reconstructs the original dimension layer by layer from the low-dimensional representation: ; in, is the decoding result; is the decoder function; is a set of decoder parameters; A regularization term is added to the loss function to prevent overfitting, and the improved loss function is as follows: ; in, is the total loss; is the reconstruction error; and They are the L1 regularization coefficient and the L2 regularization coefficient respectively; is the L1 norm of the model parameters; is the L2 norm of the model parameters; During anomaly detection, the reconstruction error of each data point is calculated. If the error exceeds the dynamically adjusted threshold, it is determined to be an anomaly.
6. According to claim 1, a method for intelligent control of power distribution stations based on deep fusion of digital and analog, It is characterized in that The S5 is specifically: S51: Based on the three-dimensional digital model of the distribution station, construct a distribution station directed graph G=( ): It represents all locations in the station, including equipment locations and intermediate locations that are passed by during inspections; E represents the path between nodes; Each edge There is a weight , indicating a node Calculate the distance between S52: Use Dijkstra algorithm to calculate the shortest path; S53: Dynamically adjust the inspection path using equipment abnormalities and importance levels, by introducing an adaptive adjustment factor into the shortest path algorithm: When the device is normal, the path weight is the distance plus the importance adjustment factor: ; If the node If an exception occurs, the path weight will increase by an exception weight. : ; in, Representation Node The physical distance between Representation Node level of importance; Representation Node Abnormal situation, 0 means no abnormality, 1 means abnormality; is the importance modifier; It is an abnormal regulatory factor.
7. According to claim 6, a method for intelligent control of power distribution stations based on deep fusion of digital and analog, It is characterized in that The Dijkstra algorithm is used to calculate the shortest path, as follows: Set the initial distance for all nodes v , set dist(s)=0 for the starting node s, create a priority queue, and add the starting node to the queue; Extract the node u with the smallest distance from the priority queue. If node u has not been processed yet, mark it as processed; For Node , For all neighbor nodes of node u, calculate the distance from the starting node s to the node The path distance dist( ), if the path through u is shorter, update dist( ) and add it to the priority queue; The above steps are repeated until all nodes have been processed or the target node has been reached.
8. According to claim 1, a method for intelligent management and control of power distribution stations based on deep fusion of digital and analog, It is characterized in that The step S6 is specifically as follows: S61: Real-time acquisition of environmental data through the laser radar, IMU and camera configured on the inspection robot; S62: Utilize SLAM algorithm to continuously update the three-dimensional map of the power distribution station during navigation; S63: During the path planning process, the known position of the 3D model and the map constructed by SLAM in real time are combined to correct the current position of the robot and adjust the inspection path in real time.
9. According to claim 8, a method for intelligent control of power distribution stations based on deep fusion of digital and analog, It is characterized in that The SLAM algorithm is used to continuously update the three-dimensional map of the power distribution station during the navigation process, as follows: The robot state is integrated into several particles, each particle includes its own position state and the corresponding map; According to the control input u t Make state predictions for each particle and add noise: in, Represents particles In time Status; Indicates at time The control input of f() is the motion model function; Represents particles In time The weight of Update the map information in each particle and calculate the likelihood between the map and the observation: Resample according to the likelihood value and keep the most likely particles: ; ; in, Represents particles In time Maps; Indicates that in the particle state and map Next, observe probability; Represents resampling.
